Our goal is to determine the value added of regional and longitudinal brain structural, perfusion, and diffusion tensor MRI, to predict future cognitive decline/Alzheimer's disease (AD). During the previous grant period we recruited and followed longitudinally 145 subjects scanned at 1.5T with mild cognitive complaints and impairments. We found that even after accounting for baseline memory function, memory decline was significantly predicted by entorhinal cortex volume, fractional anisotropy of the parahippocampal cingulum white matter, and perfusion of the posterior cingulate gyrus. These important new findings confirmed our original a priori hypothesis, and also demonstrated the value of DTI which was added on after the grant was funded. We also developed new methods and performed pilot MRI studies at 4 T and found that mild cognitive impairment is associated with atrophy of specific hippocampal subfields. Diffusion tensor imaging of cingulum white matter and perfusion of posterior cingulate also provide sensitive measures to detect early AD. We propose to follow subjects already enrolled at 1.5 T and also study 240 new nondemented elders at increased risk for cognitive decline with longitudinal 4T MRI including: high-resolution measurements of hippocampal subfields, diffusion spectral imaging (DSI) and perfusion MRI. Predictors will be obtained from MR at baseline and 12 months. Outcomes will be cognitive decline (change from baseline of memory function) assessed annually for the duration of the study. Our primary hypothesis is that after baseline cognition is accounted for, we can predict decline of memory function using a combination of: volume loss in CA 1-2 transition zone (a hippocampal subfield found to show significant changes in MCI), reduced fractional anisotropy of the cingulum white matter tract, and posterior cingulate perfusion. We will test whether the combined multiple correlations between the selected set of variables and the outcome are significant. We will also perform a series of structured explorations to test our biological model by 1) examining correlations and other predictors not included in the primary hypotheses and outcomes, 2) determining correlations between: hippocampal subfields, FA of parahippocampal cingulum, posterior cingulate perfusion, other MRI measures and memory, 3) exploring the prediction of imaging for decline of non-memory cognition, 4) exploring the added predictive value of longitudinal imaging. The proposed project is unique because we will use: quantification of hippocampal subfields, diffusion spectrum imaging and 3D perfusion MRI (accounting for transit time) to determine the pathophysiology of the prodromal stages of AD. Additional significance of this project is that optimum methods to predict cognitive decline/dementia will lead to improved detection of early AD, and sensitive outcome measures for treatment and prevention trials, helping ultimately to treat and prevent AD.